Community Food Environments, exploring a geomatic vision of regional scale

• Published in: 2022 IEEE Biennial Congress of Argentina (ARGENCON) pp. 1 - 8
• Main Authors: Campero, Micaela Natalia, Victoria Marinelli, Maria, Scavuzzo, Carlos Matias, Dolores Roman, Maria
• Format: Conference Proceeding
• Language: English
• Published: IEEE 07-09-2022
• Subjects: Correlation; Environment and public health; Libraries; machine learning; Obesity; open data; remote sensing; Sociology; Software; Urban areas; Vegetation mapping
• DOI: 10.1109/ARGENCON55245.2022.9940098

It is known that characteristics of community food environments (CFEs) affect the health of populations. At present, no literature supports an effective and efficient method to approach the study of CFEs at both local and regional scales in a comprehensive manner. In Latin America, research is focused the most on the study of food supply stores and their relationship with health outcomes, generally at the local scale. This paper aims to present a complete review and proposal of the change of scale of CFEs problem, taking as an example the geospatial analysis of the food environment of localities with more than 5000 inhabitants in the province of Cordoba. According to 2010 National Census, 95 localities were selected in Cordoba with a population greater than or equal to 5000 inhabitants. Obesity data were extracted from the first National Nutrition and Health Survey (2005). Based on the state-of-the-art, some indicators were selected to characterize the three dimensions (Political, Environmental, and Individual) of CFEs proposed by Glanz et al. [1]. Pearson´s correlation between indicators of CFE was calculated in R software (p< 0.05). Based on the approach proposed by Alvarez DiFino et al. [2], maps were developed in RGB colors to show the conformation of CFEs based on their three dimensions. Finally, an XGBoost model was run to identify the most influential variables on obesity, and to determine the relative importance of each variable within the model, SHapley Additive exPlanations library was used. A mild positive correlation was observed between the number of preschoolers with obesity and socioeconomic status (r=0.30) and a mild negative correlation between the number of obese preschoolers and health vulnerability index (r=-0.30). The generated RGB map shows clearly that the south of the province (localities with higher importance of individual dimension) showed a lower number of cases of obesity compared to the north- northwest of the territory (cities with a higher environmental component and a lower political dimension). Variables and their values, most influential in the obesity modelling were the higher average vegetation index, followed by a lower level of chronic poverty and a higher level of risk of food insecurity. The methods applied to study the CFEs in Latin America are not replicable on a large scale. Our proposal is feasible and shows great potential, considering not only the use of resources but also the incorporation of factors involved in the genesis of a health problem to facilitate efficient decision-making.